25814-P.R.(109-118)chaos.

Fnl 8/8/00 9:16 AM Page 109

Strategy at the
edge of chaos
Eric D. Beinhocker

“Fishbowl” economics once provided the basis of corporate strategy, but

no longer. New theories show that markets are “complex adaptive systems.”
Can managers be more than blind players in an evolutionary business game?

T he economist Paul Krugman says that there are three types of

economics: up-and-down economics (“stocks were up and unemploy-
ment was down today”), airport-bookstore economics (Ten Easy Steps to
Avoid Global Depression), and Greek-letter economics. Greek-letter eco-
nomics is the mathematical variety, practiced in universities and published
in academic journals. And it is in serious trouble.

Historically, Greek-letter economics has rewarded mathematical pyrotechnics

over fidelity to the real world. The core theories that Greek-letter economics
has produced over the last few decades, such as “rational expectations”
and “general equilibrium” theory, are mathematically elegant but lacking in
empirical validation.

The author thanks Kevin Coyne, Oliver Engert, Dick Foster, Rainer Gawlick, John Hagel, Somu Subramaniam,
George Ugras, and the author’s colleagues in the McKinsey Strategy Forum/Strategy Theory Initiative for
their contributions to this article. Mr. Beinhocker also benefited from the writings of, and discussions with,
W. Brian Arthur, Larry Blume, Steven Durlauf, David Lane, and Stuart Kauffman of the Santa Fe Institute.
Eric Beinhocker is a principal in McKinsey’s Washington, DC, office. This article was originally pub-
lished in The McKinsey Quarterly, 1997 Number 1. Copyright © 1997 McKinsey & Company. All rights
reserved.
This article can be found on our Web site at www.mckinseyquarterly.com/strategy/sted97.asp.

109
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 110

110 S T R AT E G Y I N T H E N E W E C O N O M Y

The dismal state of the dismal science matters to managers, chief operating
officers, consultants, and business professors because much of modern
management thinking has been built on a foundation of Greek-letter
economics. The bad news is that this foundation is now in serious doubt.
But take heart — the good news is that a radically new one is being put
in place.

The roots of management thinking

Many of the most successful and widely used strategy tools today—the
five-forces framework, cost curves, the structure-conduct-performance
(SCP) model, and the concept of sustainable competitive advantage, to
name a few— owe their origins to ideas developed in the 1950s in a field
known as the theory of industrial organization. Industrial organization
theory, which is concerned with industry structure and firm performance,
is in turn based on microeconomic theory.

Modern neoclassical microeconomics was founded by Leon Walras,

William Stanley Jevons, and Carl Menger in the 1870s and synthesized
into a coherent theory by Alfred Marshall at the turn of the last century.
Seeking to make economics more scientific, Walras, Jevons, and Menger
borrowed ideas and mathematical apparatus from the leading science of
their day: energy physics. They copied the mathematics equation by equa-
tion, translating it metaphorically (and, according to many physicists,
incorrectly) into economic concepts.1

In the mid-19th century, energy physicists developed a theory of closed

equilibrium systems, which provides the core metaphor of Alfred Marshall’s
traditional economics and much of today’s management thinking. Consider
a ball at the bottom of a bowl. If no energy or mass enters or leaves the
bowl— that is, if the system is closed— the ball will sit in equilibrium at its
bottom forever. In economic terms, the sides of our bowl represent the struc-
ture of a market (for instance, producer costs and consumer preferences),
and the gravity that pulls the ball to its lowest energy state represents profit-
seeking behavior, pulling a firm to its highest-profit state. If we know the
economic forces at work, and if firms are rational, we can predict where the
ball will come to rest in the bowl—in other words, the prices, quantities
produced, and profitability of firms under equilibrium. If some exogenous
shock hits the system (if, say, a technology shift alters producer costs), the
sides of the bowl change shape, and the ball rolls to a new point of equilib-
rium (Exhibit 1).
1
For an account of the interwoven history of economics and physics, see Philip Mirowski, More Heat than
Light: Economics as Social Physics, Physics as Nature’s Economics, Cambridge: Cambridge University
Press, 1991.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 111

S T R AT E G Y AT T H E E D G E O F C H A O S 111

In a typical modern strategic analysis, a company looks at its position in

the current industry structure, considers the shocks and changes that are
occurring or might occur, and then develops a point of view on how the
industry is likely to change and what that means for its own strategy. Such
an approach makes three important assumptions: that the industry struc-
ture is known, that diminishing returns apply, and that all firms are per-
fectly rational. But what happens if rapid technological or business system
innovation makes producer costs and consumer preferences uncertain, as is
increasingly the case in today’s dynamic, high-tech, and service-dominated
economy? What if we face not diminishing returns (where each additional
acre of soybean planted is on poorer land and thus yields a lower return),
but increasing returns (where each extra Netscape browser sold increases
the value of the World Wide Web and thus yields a higher return)?2 What
if firms lack complete information, or different firms interpret the same
information in different ways?

Should the fundamental assumptions underlying the equilibrium model

be relaxed, the effect on the ball in the bowl will be dramatic. The sides
of the bowl start to bend and flex, losing their smooth shape and becoming
a landscape of hills and valleys; and the ball can no longer tell which way
is up. Now it is impossible to predict where the ball will roll, and Alfred
Marshall loses his equilibrium. And this is not merely a theoretical problem,
since the ball-in-the-bowl equilibrium model is the basis for our ideas on
strategy.

Complex adaptive systems

Anthills are marvelous things. With elaborate labyrinths of tunnels, layouts
reflecting their occupants’ social hierarchies, chambers dedicated to specific

EXHIBIT 1

Equilibrium systems

The ball rests at its lowest An exogenous shock ...and reshapes the bowl,
energy state in equilibrium. unsettles the system... creating a new equilibrium.

2
For a more detailed discussion of increasing returns, see W. Brian Arthur, “Positive feedbacks in the
economy,” The McKinsey Quarterly, 1994 Number 1, pp. 81–95; and “Increasing returns and the new
world of business,” Harvard Business Review, July–August 1996, pp. 100– 9.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 112

112 S T R AT E G Y I N T H E N E W E C O N O M Y

functions, and carefully sited entrances and exits, they are as thoughtfully
constructed as any condominium complex. Yet who is the engineer? Where
is the blueprint?

The answer, of course, is that none exists. The plan for the anthill does not
reside in any individual ant. Rather, each ant is programmed by its DNA to
obey a set of relatively simple rules,
such as “stand between two other
Individually, ants don’t do much. ants and pass along anything that is
But put them in an interactive handed to you.” Ants communicate
group and an anthill emerges with each other via chemical signals
known as pheromones, which pro-
vide inputs and outputs for the rules
and switch them on and off. It is the dynamic interaction of the rules and
signals that creates the anthill structure.

An ant community is an example of a complex adaptive system.3 Such systems

share the following three characteristics.

First, they are open, dynamic systems. The Marshall ball-in-a-bowl system
is closed; no energy or mass enters or leaves, and the system can settle into
an equilibrium state. By contrast, the energy and mass that constantly flow
through a complex adaptive system keep it in dynamic disequilibrium. An
anthill, for example, is a perpetual-motion machine in which patterns of
behavior are constantly shifting; some patterns appear stable, others chaotic.

Second, these systems are made up of interacting agents, such as ants, people,
molecules, or computer programs. What each agent does affects one or more
of the other agents at least some of the time; this creates complexity and
makes outcomes difficult to predict. The interactions of agents in a complex
system are guided by rules: laws of physics, codes of conduct, or economic
imperatives such as “cut prices if your competitor does.” If the repertoire of
rules is fixed, the result is a complex system. If the rules are evolving, as with
genes encoded in DNA or the strategies pursued by players in a game, the
result is a complex adaptive system.

Third, complex adaptive systems exhibit emergence and self-organization.

As individuals, ants don’t do much. But put them in a group where they can
interact, and an anthill emerges. Because the anthill rises out of the bottom-up

3
For an excellent nontechnical introduction to the field of complex adaptive systems, see M. Mitchell
Waldrop, Complexity, New York: Simon & Schuster, 1992; Stuart A. Kauffman, At Home in the Universe,
New York: Oxford University Press, 1995; and John H. Holland, Hidden Order, Redwood City, CA: Addison-
Wesley, 1995. For a more technical overview of complexity and economics, see Philip W. Anderson,
Kenneth J. Arrow, and David Pines (eds.), The Economy as an Evolving, Complex System, Redwood City,
CA: Addison-Wesley, 1988; and W. Brian Arthur, David Lane, Steven N. Durlauf (eds.), The Economy as an
Evolving, Complex System II, Redwood City, CA: Addison-Wesley, 1997.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 113

S T R AT E G Y AT T H E E D G E O F C H A O S 113

dynamic interactions of the ants and not from a top-down master plan, it is said
to be self-organized. The emergent structure is independent of specific agents;
while individual ants may come and go, the pattern of the anthill persists.

Other examples of complex adaptive systems include cities, forest ecosys-

tems, the immune system, and the Internet. Over the past 20 years—aided
by advances in mathematics, physics, chemistry, and biology and by the wide
availability of cheap computing power—scientists have begun to find that
complex adaptive systems are governed by deep common laws. Just as the
laws of statistics can help us understand such ostensibly different systems as
poker games and the spread of disease in populations, the laws of complexity
may yield new insights into problems ranging from the origins of life to
traffic jams in Los Angeles.

The new economics

To managers, this should be more than just interesting science. A number
of economists are beginning to say that economies are complex adaptive
systems. The case has yet to be proved, but there is circumstantial evidence
as well as support from some eminent economists, among them Kenneth
Arrow, a Nobel Prize winner and one of the prime architects of the modern
neoclassical model, and Brian Arthur of the Santa Fe Institute. (Indeed, the
new economics is sometimes referred to as the Santa Fe school of economics,
after the interdisciplinary research center with which many economists
working on ideas of complexity are affiliated.)

A new economics based on complex adaptive systems is still in its infancy, but
enough work has been done to suggest what the key components might be.

Wisdom

First, the new economics will be based on a realistic model of cognitive

behavior. Traditional economics assumes that people are alike in their
thought processes (though their preferences may differ) and that they make
choices as if they were solving complicated deductive equations that would
enable them to make the best possible decisions. Economists have long real-
ized that these assumptions are too simple, but such assumptions were
needed to make the math work for the ball-in-the-bowl model. Computer
simulation techniques and advances in cognitive science now allow econo-
mists and others to make much more realistic assumptions about people’s
decision-making processes—assumptions based on skills such as the ability
to recognize patterns and to develop inductive rules of thumb by learning
from experience. Early work suggests that replacing perfect rationality with
more realistic assumptions about inductive, nonoptimal decision making
produces new insights and different strategic recommendations.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 114

114 S T R AT E G Y I N T H E N E W E C O N O M Y

Webs

Second, the new economics will see agents interacting with one another in
a dynamic web of relationships. It will not be enough to have a sound model
of a firm’s behavior; you must also know how people interact within the firm,
how it interacts with other firms in its market, and how these interactions
change over time.

Waves

Third, markets will be viewed as inherently dynamic rather than static systems,
thus making possible an explanation of growth and innovation that traditional
economics has never been able to provide. When adaptive agents interact in a
web of relationships, evolutionary changes in one agent affect the evolution of
others. This effect, known as coevolution, is frequently seen in nature and also
occurs in economics when an innovation (such as the invention of the automo-
bile) produces ripple effects throughout the whole economy (the development
of the oil industry, motels, the growth of suburbs, and so on).4

Would-be worlds

Not only the substance but also the research techniques of economics will
be transformed. Although the new economics will continue to make use of
traditional mathematical proofs, it will increasingly turn to sophisticated
computer simulations based on more realistic assumptions. In agent-based
models, for example, a company can be modeled as an intelligent computer
program capable of learning and adapting. You can put a set of these pro-
grams into a simulated competitive market, unleash the forces of evolution
on them, and watch different futures unfold. Since complex systems can
be difficult or impossible to forecast, such models will be of little help in
forecasting the precise path an industry might take; however, they will be
valuable in helping to determine how and why markets behave as they do.

Early thinking on management

The new economics has advanced far enough for us to begin to make pre-
liminary hypotheses about its implications for strategy and organization.
One characteristic of complex adaptive systems is punctuated equilibrium
(Exhibit 2). This natural endogenous feature of the evolutionary process
occurs when times of relative calm and stability are interrupted by stormy
restructuring periods, or “punctuation points.” Punctuated equilibrium
makes it difficult for participants to survive for long periods, as their strate-
gies and skills tend to get finely optimized for the stable periods and then
4
See Stuart A. Kauffman, “Technology and evolution: Escaping the Red Queen effect,” The McKinsey
Quarterly, 1995 Number 1, pp. 118–29.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 115

S T R AT E G Y AT T H E E D G E O F C H A O S 115

suddenly become obsolete when the inevitable restructuring takes place.

Similarly, companies have a hard time surviving the upheavals, shakeouts,
and technology shifts that punctuate the evolution of markets. To prosper
in the long run, a company must adapt as readily as its market, or more so.
More specifically, it must be both a strong competitor in the current regime
and a smart evolver, able to innovate ahead of the market or to adapt with it.

The equilibrium view of strategy has focused on how to be a good com-

petitor; let us consider five critical aspects of being a good evolver.

Focused versus robust strategies

Traditional strategy tends to emphasize a single focused line of attack—a

clear statement of where, how, and when to compete. In a complex adaptive
system, a focused strategy to dominate a niche is necessary for day-to-day
survival but not sufficient in the long run. Given an uncertain environment,
strategies must also be robust—that is, capable of performing well in a variety
of possible future environments.5

Competitive advantage versus continuous adaptation

Evolutionary systems exhibit a phenomenon known as the Red Queen effect,

after that character’s remark in Through the Looking Glass: “It takes all the
running you can do to keep in the same place.”6 In nature, the Red Queen
effect is at work when a predator learns to run faster; its prey responds by
acquiring better camouflage; the predator then develops a better sense of smell;
the prey starts to climb trees; and so on. Evidence suggests that the business
world resembles a Red Queen race.
EXHIBIT 2
A study of the performance of more
than 400 companies over 30 years Punctuated equilibrium
reveals that firms find it difficult
to maintain higher performance Punctuation points
levels than do their competitors for
more than about five years at a time
(Exhibit 3, on the next page). Long-
Change

term superior performance is achieved

not through sustainable competitive Stasis

advantage but by continuously devel-

oping and adapting new sources of
temporary advantage and thus being
Time
the fastest runner in the race.

5
For an in-depth discussion of this topic, see Eric D. Beinhocker, “On the origin of strategies,” on page 167
of this anthology.
6
See Stuart A. Kauffman, “Technology and evolution: Escaping the Red Queen effect,” The McKinsey
Quarterly, 1995 Number 1, pp. 118–29.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 116

116 S T R AT E G Y I N T H E N E W E C O N O M Y

EXHIBIT 3

The Red Queen effect

Industry mean
Total return to shareholders, percent

1 2 3 4 5 1 2 3 4 5 1 5 10 15 20
Years Years Years
If advantage was sustainable, Instead, a study of 400 companies Although companies do not sustain
some companies would be over 30 years shows that high advantage, a few seem to create
expected to sustain superior performers regress to the industry new advantages continually.
performance. mean after three to seven years.

Conservative operator versus radical innovator

In a complex adaptive system, an agent that is resistant to change and not

adaptable will have low fitness, and so, conversely, will an agent that is
oversensitive to shifts in its environment and constantly making radical
responses. But between these extremes of stasis and chaos lies a region—
the edge of chaos—where fitness is maximized (Exhibit 4). Being at the
edge of chaos means
EXHIBIT 4
something more subtle
The edge of chaos than pursuing a mod-
erate level of change. At
High
the edge of chaos, one
is simultaneously con-
Level of fitness

servative and radical.

Evolution is adept at
keeping things that
Low work while at the same
Stasis Edge of chaos Chaos
time making bold exper-
iments. The morphology
Slow, incremental change Rapid, radical change
of the spinal cord is a
robust adaptation that
has survived eons of
evolution and enormous environmental shifts. Yet nature has experimented
wildly around this core idea, producing vertebrates that range from birds to
whales to humans.

The Walt Disney Company is a firm that prospers at the edge of chaos. Its
theme parks and other businesses are run in a deeply conservative fashion.
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 117

S T R AT E G Y AT T H E E D G E O F C H A O S 117

A strong culture supports Disney’s mission of providing family entertain-

ment. In operations, no detail is too small, right down to the personal
grooming of the parking-lot attendants. This culture is ingrained in the
organization and constantly reinforced through management processes.

At many organizations, such a conservative culture and such tightly con-

trolled operations would snuff out creativity. Yet Disney manages to be one
of the most innovative companies in the world. It pioneered animated films
and destination theme parks, built EPCOT, linked media and retail with its
Disney Stores, and took an early lead in cable television. Disney manages
the tension between conservatism and innovation by maintaining an almost
cultlike attention to detail and discipline but at the same time forgiving
honest mistakes made in the pursuit of innovation.

Routinized versus diverse strategies

Another requirement for success in evolutionary systems is a rich pool of

possible strategies. This diversity represents the source of the innovations
that keep a player ahead in the Red Queen race and can be drawn on to
develop responses when the environment changes. But diversity also has its
cost. Many mutations are harmful and selected against, limiting the diversity
found in a species. Moreover, a certain level of standardization is beneficial:
a relatively narrow range of mating behavior, for example, probably makes
it easier for interested parties to find one another. Evolution strikes a bal-
ance, standardizing designs that work but seeding the population with
enough variation to provide a basis for future innovation and adaptation.

Few companies are skilled at striking this balance. The result is firms that
are either chaotic or vulnerable at punctuation points because they no longer
have a well-stocked pool of ideas and experiences from which to draw.

Scale versus flexibility

In traditional strategic and organizational thinking, big is good. Benefits of

scale are easy to identify in purchasing, operations, marketing, and so on.
Why is it, then, that big companies can have such a hard time responding to
attacks by smaller competitors? A complexity-based view can shed light on
the downside of size.

A simple system with relatively few parts and interconnections isn’t highly
adaptable: the number of states it can manifest is small compared with the
number of situations it might encounter. As the system grows bigger and
more complex, the number of states it can manifest, and thus its repertoire
of possible responses to changes in its environment, grows exponentially.
However, beyond a certain level of scale and complexity, its adaptiveness
25814-P.R.(109-118)chaos.Fnl 8/8/00 9:16 AM Page 118

118 S T R AT E G Y I N T H E N E W E C O N O M Y

drops off rapidly in what Stuart Kauffman calls a complexity catastrophe.

This occurs when the epistasis, or interaction between the parts, builds to
such an extent that any positive change in one part has ripple effects that
cause negative changes elsewhere. The system thus becomes more conserva-
tive as it grows, and finding adaptations that don’t have harmful side effects
gets harder and harder.

When Dell Computer began to do well at selling inexpensive personal com-

puters by mail, no doubt someone at IBM said, “Why don’t we do that too?”
But IBM couldn’t follow suit without damaging its extensive distribution
channels of dealers and direct salespeople. Its history and size created a
trade-off that Dell didn’t face and made it difficult for IBM to respond.

Companies can mitigate the effect of complexity catastrophes through strategic

and organizational changes. GM started Saturn in a greenfield organization
precisely to free it from the constraints of corporate bureaucracy. AT&T split
itself in three to create smaller organizations and reduce strategic conflict.

Becoming competitors and evolvers

As a complexity-based view of economics develops, new tools will be devised
to help managers fashion better-evolving companies. Some of these tools
will be analytical: options theory and evolutionary modeling to help develop
robust strategies, for example. Others will be conceptual: new organizational
forms that help avert complexity catastrophes, say, or practices that promote
a rich fund of ideas.

Becoming a better evolver will be a major challenge for most companies; it is

difficult enough just to be a successful competitor. And how do you motivate
a thriving organization in a stable regime to take on the task of becoming
more innovative and adaptive so as to meet challenges it can’t even foresee?
Equally, a company struggling through a major punctuation point finds it
hard to worry about its long-term evolvability.

But for companies that do accept the challenge, the payoff promises to be
considerable. Unlike creatures in nature, we are not blind, passive players in
the evolutionary game. Through the sciences of complexity, we can come to
understand how evolution works, the tricks it has up its sleeve, and the skills
needed to survive in a complex world. If we do so, we may be able to har-
ness one of the most powerful forces of all: evolution will then be the wave
we ride to new levels of creativity and innovation rather than the tide that
washes over us.